The CMEMS Globcolour Chlorophyll a Product Based on Satellite Observation: Multi-Sensor Merging and ﬂagging Strategies

Ocean Sci., 15, 819–830, 2019 https://doi.org/10.5194/os-15-819-2019 © Author(s) 2019. This work is distributed under the Creative Commons Attribution 4.0 License.

The CMEMS GlobColour chlorophyll a product based on satellite observation: multi-sensor merging and ﬂagging strategies

Philippe Garnesson, Antoine Mangin, Odile Fanton d’Andon, Julien Demaria, and Marine Bretagnon ACRI-ST, Sophia Antipolis, 06904, France

Correspondence: Philippe Garnesson ([email protected])

Received: 21 December 2018 – Discussion started: 2 January 2019 Revised: 20 May 2019 – Accepted: 21 May 2019 – Published: 24 June 2019

Abstract. This paper concerns the GlobColour-merged will require a better characterisation and additional inter- chlorophyll a products based on satellite observation (SeaW- comparison with in situ data. iFS, MERIS, MODIS, VIIRS and OLCI) and disseminated in the framework of the Copernicus Marine Environmental Monitoring Service (CMEMS). This work highlights the main advantages provided by the 1 Introduction Copernicus GlobColour processor which is used to serve CMEMS with a long time series from 1997 to present at The Copernicus Marine Environmental Service (CMEMS) the global level (4 km spatial resolution) and for the Atlantic provides regular and systematic reference information on the level 4 product (1 km spatial resolution). physical state and on marine ecosystems for global oceans To compute the merged chlorophyll a product, two major and for European regional seas, including temperature, cur- topics are discussed: rents, salinity, sea surface height, sea ice, marine optical properties and other such parameters. – The first of these topics is the strategy for merging This capacity encompasses satellite and in situ data- remote-sensing data, for which two options are consid- derived products, the description of the current situation ered. On the one hand, a merged chlorophyll a product (analysis), the prediction of the situation a few days ahead computed from a prior merging of the remote-sensing (forecast) and the provision of consistent retrospective data reflectance of a set of sensors. On the other hand, a records for recent years (re-analysis). merged chlorophyll a product resulting from a combi- The Ocean Thematic Assembly Centre (OCTAC) is part nation of chlorophyll a products computed for each sen- of CMEMS and is dedicated to the dissemination of ocean sor. colour (OC) products derived from satellite-based remote – The second topic is the flagging strategy used to discard sensing (Le Traon et al., 2015). OCTAC provides global and non-significant observations (e.g. clouds, high glint and regional (Arctic, Atlantic, Baltic Sea, Black Sea and Mediter- so on). ranean Sea) products for the period spanning from 1997 to the present. These topics are illustrated by comparing the CMEMS Glob- For global products, the Copernicus GlobColour proces- Colour products provided by ACRI-ST (Garnesson et al., sor has been used operationally since 2009 to serve CMEMS 2019) with the OC-CCI/C3S project (Sathyendranath et al., and its precursors (a series of EU research projects called 2018). While GlobColour merges chlorophyll a products MyOcean). with a specific flagging, the OC-CCI approach is based on The GlobColour processor was initially developed in the a prior reflectance merging before chlorophyll a derivation framework of the GlobColour project that started in 2005 as and uses a more constrained flagging approach.Although an ESA Data User Element (DUE) project to provide a con- this work addresses these two topics, it does not pretend tinuous data set of merged L3 ocean colour products. Since to provide a full comparison of the two data sets, which the beginning of the project, GlobColour has been continu-

Published by Copernicus Publications on behalf of the European Geosciences Union. 820 P. Garnesson et al.: The CMEMS GlobColour chlorophyll a product Based on satellite observation ously used by more than 600 users worldwide. This effort has been continued in the framework of CMEMS to derive (among others) the chlorophyll a ocean colour core product. Many algorithms have been published to retrieve chlorophyll a from reflectance data derived from satellite observations (e.g. Müller-Karger et al., 1990; Aiken et al., 1995; Morel, 1997; O’Reilly et al., 1998, 2000). Since early 2018, the CMEMS GlobColour chlorophyll a product has been based on the merging of three algorithms: – the CI approach (Hu et al., 2012) for oligotrophic water (70 % to 80 % of ocean), – the common approach OCx (OC3, OC4 or OC4Me depending on the sensor) for mesotrophic water and – the OC5 algorithm (Gohin et al., 2002) for complex waters, which is of specific interest for end users who man- age complex waters along the coastal zone. It should be noted that the OC5 algorithm is based on a lookup table implementation that handles both complex and mesotrophic waters (see Sect. 2.1) The work presented here highlights the conceptual advantage of the CMEMS Copernicus GlobColour processor with regards to the flagging and merging of sensors. In the following sections, results are described and illustrated with comparison to the OC-CCI products. The comparison between GlobColour and OC-CCI is especially relevant as the same chlorophyll a algorithms (CI and OC5) are used by the two initiatives: this implies that the differences in performance mainly result from the merging strategies and flagging schemes. Figure 1. Swath of the different sensors used at present by CMEMS for (a) MODIS Aqua, (b) VIIRS-NPP and (c) OLCI-S3A. In prac- 2 Methods tice the effective swath coverage is reduced, which is mainly due to clouds or sun glint effects. Source: http://octac.acri.fr (last access: 2.1 Merging approach 15 January 2019).

At present, the CMEMS GlobColour-merged chlorophyll a product relies on the following sensors: SeaWiFS (1997– All of the sensors used observe the Earth along a helio- 2010), MERIS (2002–2012), MODIS Aqua (2002–present), synchronous orbit. Figure 1 displays the coverage of a single VIIRS-NPP (2012–present) and OLCI-S3A (2016–present). sensor that is unable to provide the full Earth coverage over a The long time series from 1997 to the present relies on dif- single day (Maritorena et al., 2015). VIIRS provides a larger ferent sensors, observing the Earth at different spectral bands swath than the other sensors, but the coverage is incomplete (and different bandwidths), with different acquisition times due to sun glint. (so different atmospheric and sun conditions), and with dif- When more than one sensor is available for the same pe- ferent spatial resolutions from about 300 m to 1 km at nadir riod it is possible to take advantage of their complementarity (larger on the swath border). The main characteristics of the and redundancy for a number of benefits. For instance, as sensors/bands used for CMEMS are summarised in Table 1. they may record the same spot at different times of the day, VIIRS-NOOA20 and OLCI-S3B will be ingested in the op- morning haze may impact one sensor but not another (Toole erational products in 2019. et al., 2000). It should be noted that the global chlorophyll a product To compute a multi-sensor chlorophyll a product, two is provided at a 4 km spatial resolution at present, but the main merging approaches exist: objective in the coming years is to provide (at least along the coast) a chlorophyll a product at a 300 m resolution. – The first approach (used by OC-CCI) is based on merged reflectance of remote-sensing (RRS) computed

Ocean Sci., 15, 819–830, 2019 www.ocean-sci.net/15/819/2019/ P. Garnesson et al.: The CMEMS GlobColour chlorophyll a product Based on satellite observation 821

Table 1. Main characteristics of sensors/bands used for CMEMS (VIIRS-JPPS1 and OLCI-S3B will be used by the GlobColour processor in the framework of the CMEMS release scheduled in July 2019).

Sensor RRS wavelengths (nm) Spatial resolution Swath Equator Period at nadir (km) width (km) crossing time SeaWiFS 412, 443, 490, 510, 555 and 670 1 and 4 1502 12:20 1997–2010 MERIS 413, 443, 490, 510, 560, 620, 1 and 0.3 1150 10:00 2002–2012 667, 681 and 709 MODIS Aqua 412, 443, 469, 488, 531, 547, 1 2330 13:30 2002–present 555, 645, 667 and, 678 VIIRS-NPP 410, 443, 486, 551 and 671 1 3040 10:30 2012–present OLCI S3A 400, 412, 442, 490, 510, 560, 1.2 and 0.3 1270 10:00 2016–present 620, 665, 674, 681 and 709 VIIRS-JPPS1/NOAA20 411, 445, 489, 556 and 667 0.75 3040 09:50 December 2017– present OLCI S3B 400, 412, 442, 490, 510, 560, 1.2 and 0.3 1270 10:00 2018–present 620, 665, 674, 681 and 709

in a prior step and then used to derive chlorophyll a. chlorophyll a, the input level 2 reflectance and input flags are Based on a band shifting and bias correction approach, used (level 2 provides flags/indicators regarding the quality merged RRS for the standard SeaWiFS wavelengths of the reflectance on a pixel basis). For instance, a pixel can (412, 443, 490, 510, 555 and 670 nm) are provided. The be impacted by a sun glint effect. In such cases reflectance blended chlorophyll a algorithm used in the OC-CCI data are available, but agencies recommend not to use them v3.1 release attempts to weight the outputs of the best for further processing. Each space agency publishes flagging performing algorithms based on the water types present. data based on its own strategy, which has been designed to guarantee the quality of its products. The drawback is that for – Conversely, in a second approach (used by CMEMS complex water (especially along coastal area), a large propor- GlobColour), chlorophyll a is computed in an initial- tion of the data is flagged, resulting in level 3 products with isation step for each sensor using its specific charac- limited coverage. teristics (spectral band and resolution), and then the Since early 2018, the GlobColour processor has been mod- mono-sensor chlorophyll a products are resampled and ified to apply a lower level of flagging, resulting in a better merged. The continuity of the different algorithms was spatial and temporal coverage. This modification is inher- initially obtained using a water classification approach ited from the OC5 algorithm (Gohin et al., 2002) which was (Saulquin al., 2018). Since early 2018, a new approach initially designed for coastal monitoring. OC5 uses its own has been adopted: strategy to flag data: the algorithm uses both the official flags and empirical thresholds that have been tuned for each sen- – First, the continuity of algorithms used for sor (e.g. the OC5 sun zenith angle, SZA, is set to 78◦ instead mesotrophic and complex waters is guaranteed by of 70◦). the OC5 lookup table. The OC5 lookup table is ini- The OC5 flagging strategy is used by the GlobColour tialised using the OC3 and OC4 coefficients from sensor approach but not by the OC-CCI approach, which agencies and then empirically adjusted when the uses a specific OC5 lookup table applied on the merged re- green band exceeds a given threshold (see Gohin flectances. et al., 2002, for details). Concerning OC-CCI, the flagging strategy for the v3.1 re- – Then, the CI and OC5 continuity is ensured using lease depends on the sensor. When the reflectance data orig- the same approach as that utilised by NASA. When inate from level 2 data provided by the agency (SeaWiFS the chlorophyll a concentration is in the range from and VIIRS-SNPP) the official flags are applied. When the 0.15 to 0.2 mg m−3, a linear interpolation of OC5 POLYMER algorithm (Steinmetz et al., 2011) is used for at- and CI is used. This provides continuity between mospheric correction (MERIS and MODIS) a generic pixel the two algorithms. identification and classification algorithm called Idepix is used (part of BEAM software) instead of the standard POLY- 2.2 Flagging approach MER flagging (which is too permissive).

Inputs of the Copernicus GlobColour processor are the level 2 products provided by the space agencies. To derive www.ocean-sci.net/15/819/2019/ Ocean Sci., 15, 819–830, 2019 822 P. Garnesson et al.: The CMEMS GlobColour chlorophyll a product Based on satellite observation

3 Results and discussion VIIRS has suffered from a significant drift since its launch, as illustrated by the January evolution for the years 2012, 3.1 The reflectance merging approach 2016 and 2019. In January 2012, 90 % of the MODIS pixels at global level were higher than VIIRS-SNPP, whereas in The reflectance merging approach is used by OC-CCI to de- January 2019 100 % of the VIIRS-SNPP pixels were higher rive the global chlorophyll a OC-CCI time series and by than MODIS. the regional CMEMS products. As pointed out by Volpe et Another major difficulty with merging RRS for the differ- al. (2019) for the regional Mediterranean products, the band ent sensors is that the observed bias varies according to the merging approach has the advantage of providing a homoge- region and season considered, as previously shown by the ar- neous data set of spectral reflectance from which the follow- tificial trends along the years. ing environmental parameters (amongst others) can be de- Figure 5 shows the inter-comparison of RRS at about rived, in full consistency for the long term: chlorophyll a 670 nm between VIIRS-SNPP and OLCI-S3A compared products, light attenuation, Kd and suspended particulate with MODIS. It shows very important bias (e.g. 82 % of matter. the OLCI pixels exceed a relative difference of 20 %), and However, the consistency of the long time series provided in the case of MODIS, the equatorial zone shows different by OC-CCI suffers from some limitations. Figure 2 from the behaviour to the high latitudes. OC-CCI Product User Guide inter-compares the different re- The impact of such reflectance merging on chlorophyll a leases of the OC-CCI time series. It shows that the V2 release is shown in Fig. 6. For the year 2018, limitation is visible was strongly impacted by the MERIS sensor ceasing opera- regarding the ability to handle clear water between tracks of tion in April 2012 (see Table 1). The V3 release demonstrates the different sensors (the colour scale has been set to range a trend depending on the sensors used: it increases for the pe- from 0.01 to 0.2 mg m−3). riod from 2002 to 2010, based on the contributions of SeaW- The previous illustrations demonstrate the limitation of iFS, MODIS and MERIS, and decreases for the period from the assumption of consistency along the OC-CCI time series 2012 to 2017, as only MODIS and VIIRS-SNPP were used. (and on a daily basis in Fig. 6), but it is clear that the Glob- On a regional scale, strong limitations regarding the con- Colour products are also affected by the quality of input RRS sistency of the time series and trends derived are observed as upstream. illustrated for the Arctic in Fig. 3. Strong variations are also Figure 7 shows the month to month evolution of the observed throughout the years. For instance, before 2002, chlorophyll a median concentration computed at the global only SeaWiFS is available, limiting the quality of the output level for OC-CCI and GlobColour products. To make the two data. Furthermore, the end of the MERIS lifetime and start data sets inter-comparable, monthly statistics are computed of VIIRS SNPP in 2012 causes the same trend as described on common pixels. The yellow sections on the plot show the above. change of sensor combinations. Trends and gaps should be It is known that both MODIS and VIIRS instruments have carefully interpreted according to the sensors used. The os- major calibration issues starting in about 2012. VIIRS-SNPP cillations of the median through time are linked to the sen- degradation was identified a few months after launch, and sor coverage, which moves from north to south for high lati- MODIS, while designed for a lifetime of 7 years, is still op- tudes (due to the variation of the sun zenith angle which ren- erating after 17 years. MODIS calibration has required reg- der measurements in winter unfeasible). At the beginning of ular modifications to adjust the temporal trends (R2009.1, SeaWiFS (September 1997 to mid-1998) data were only par- R2010.0 and especially R2012.0). Since the end of 2017, tially available, resulting in limited coverage. When MODIS a NASA reprocessing called R2018 has significantly allevi- and MERIS started providing data in 2002, the spatial cover- ated VIIRS issues and, more importantly, has corrected the age was improved. This means that the gap at this time does MODIS drift using a new procedure to regularly update the not correspond to a change in the continuity of the time se- MODIS calibration (available about 3 months after acquisi- ries, but to a change in the spatial coverage of pixels used to tion). compute the median. The higher values for OC-CCI are most It should be noted that this new NASA processing (called likely linked to the NASA R2018 reprocessing that is used by R2018.1) does not yet benefit the full OC-CCI series (only GlobColour, but not yet by OC-CCI. During this R2018 re- for the recent OC-CCI v3.1 extension until June 2018), but a processing, MOBY (Marine Optical BuoY; in situ data) and new OC-CCI v4 release is scheduled for 2019. SeaWiFS calibration were also improved with a resulting de- It should also be mentioned that the R2018 data set still crease in chlorophyll a of the order of 10 % for all sensors. suffers from issues which will continue to impact future reprocessing attempts. Indeed, VIIRS Rrs(443) and Rrs(488) 3.2 The chlorophyll a merging approach have increased regularly since 2012, whereas MODIS is comparatively more stable. Figure 4 shows the relative The chlorophyll a merging approach is used by the Glob- Rrs(443) difference between VIIRS and MODIS (in percent) Colour processor for the CMEMS products and is not de- based on the monthly NASA R2018 global products at 4 km. signed to solve the issues linked to the upstream data. How-

Ocean Sci., 15, 819–830, 2019 www.ocean-sci.net/15/819/2019/ P. Garnesson et al.: The CMEMS GlobColour chlorophyll a product Based on satellite observation 823

Figure 2. Comparison of the global median chlorophyll a concentration as function of time for v2.0, v3.0 and v3.1 using the monthly composite as input. Source: OC-CCI Product User Guide, release 3.1.0, 24 April 2017. The yellow shaded regions show the change of sensor combinations (see Table 1).

Figure 3. Arctic time series and trend (1997–2017) from the OC-CCI product. The time series are derived from the regional chlorophyll a reprocessed (REP) products as distributed by CMEMS which, in turn, results from the application of the regional chlorophyll a algorithms with remote-sensing reflectances provided by the ESA Ocean Colour Climate Change Initiative (ESA OC-CCI). Daily regional mean values are calculated by averaging (weighted by pixel area) over the region of interest. A fixed annual cycle is extracted from the original signal, using the “Census-I” method as described in Vantrepotte and Mélin (2009). The de-seasonalised time series is derived by subtracting the seasonal cycle from the original time series, and then fitted to a linear regression to obtain the linear trend. Source: CMEMS OMI (Ocean Monitoring Indicator) QUID (Quality Information Document) (Pardo et al., 2018). ever, the sensor approach (instead starting from merged RRS) terpolations are required to simulate the 510 band (this has crucial advantages compared with the previous approach: band is not available for VIIRS-SNPP, VIIRS-JPPS1 or MODIS; see Table 1). – When a new sensor (or a new reprocessing of an exist- ing sensor) becomes available, limited effort is required – For the CI algorithm, the GlobColour processor benefits with respect to bias correction because the bias correc- from the efforts of the space agencies with respect to tion is limited to the chlorophyll a field of the sensor adjusting the coefficients accounting for the high vari- considered. For the merged reflectance approach, the ability of the 670 band (Fig. 5) for each sensor. bias correction of five reflectance measurements and in- www.ocean-sci.net/15/819/2019/ Ocean Sci., 15, 819–830, 2019 824 P. Garnesson et al.: The CMEMS GlobColour chlorophyll a product Based on satellite observation

Figure 4. Relative difference of VIIRS and MODIS at Rrs(443) (in percent) based on the monthly NASA R2018 global products at 4 km. VIIRS has suffered from a signiﬁcant trend since its launch, as illustrated by the January monthly evolution for the years (a) 2012, (b) 2016 and (c) 2019.

– It should be noted that the chlorophyll a algorithm is on reflectance with consistent time observations. Con- applied on the level 2 sensor grid, whereas in the merg- versely, when reflectance is re-projected onto a common ing approach it is applied on the common grid that is grid, it results in mixing pixels observed with an obser- required to merge the reflectance. The use of the sen- vation shift that can reach 4 h (see Table 1) in the case of sor level 2 grid guarantees that the algorithm is applied MODIS and VIIRS-JPSS1). This consideration is of mi-

Ocean Sci., 15, 819–830, 2019 www.ocean-sci.net/15/819/2019/ P. Garnesson et al.: The CMEMS GlobColour chlorophyll a product Based on satellite observation 825

Figure 5. Relative difference between sensors [(S1 − S2) / S2] regarding monthly RRS (June 2018), for (top) MODIS Rrs(667) and OLCI Rrs(665) and (bottom) VIIRS Rrs(671) and MODIS Rrs(667). Source: these plots are part of the monitoring carried out by the OCTAC and reported on http://octac.acri.fr.

nor significance with respect to the global product that Figure 10 shows that the combination of the usage of this has a spatial resolution of 4 km but will become more flagging strategy and OLCI leads to a considerable improve- significant when the product is provided at a resolution ment of the coverage without creating additional artefacts. of 300 m. Currently, both products benefit from the latest NASA R2018 reprocessing. 3.3 The flagging approach Figure 11 shows that, in certain cases, the OC-CCI coverage may be better than the GlobColour coverage. However, When compared to the official agencies’ recommendations, in this example, the OC-CCI approach is affected by signifi- the OC5 flagging strategy significantly improves the spatial cant noise, potentially due to cloud contamination. This noise coverage of the product, especially for NASA sensors. In the might be due to level 2 inputs. Indeed, while GlobColour framework of CMEMS we estimated that at the sensor level uses the level 2 product from agencies, OC-CCI starts from the coverage is increased by a factor of 3.2 for VIIRS-NPP, level 1, and applies the POLYMER algorithm to MERIS and 2.6 for MODIS Aqua, 1.6 for MERIS, 2 for SeaWiFS and 1.3 MODIS along with a specific flagging. for OLCI-S3A. Therefore, for the merged product, GlobColour chlorophyll a coverage is improved by a factor of 2.8 compared 4 Conclusion with OC-CCI. For the period from 2002 to 2012 the increase in coverage is limited to a factor of 1.5. Figures 8 and 9 show This work presents different ways to merge sensors and dif- the result of the flagging strategy depending the OC-CCI or ferent flagging strategies to estimate the daily chlorophyll a GlobColour approach. field. Compared with the chlorophyll a merging approach, the key advantage of the reflectance merging approach is that it www.ocean-sci.net/15/819/2019/ Ocean Sci., 15, 819–830, 2019 826 P. Garnesson et al.: The CMEMS GlobColour chlorophyll a product Based on satellite observation

Figure 6. Inter-comparison of the (a) OC-CCI and (b) GlobColour products (13 June 2018) for oligotrophic water (CI algorithm). The colour scale ranges from 0.01 to 0.2 mg m−3: discontinuities between tracks of the sensor clearly appear at the top of the OC-CCI case.

Figure 7. Comparison of the global median chlorophyll a concentration as function of time for OC-CCI v3.1 (blue line) and GlobColour R2018.11 (red line) at the global level and for common pixels. The yellow shaded sections show the change of sensor combinations. The discontinuities and trends of the median with time should be interpreted carefully according the sensors used.

Ocean Sci., 15, 819–830, 2019 www.ocean-sci.net/15/819/2019/ P. Garnesson et al.: The CMEMS GlobColour chlorophyll a product Based on satellite observation 827

Figure 8. The chlorophyll a concentration (15 December 2017) for (a) the OC-CCI level 3 product and (b) the GlobColour product. The two initiatives used MODIS Aqua, and VIIRS-SNPP at this time, while OLCI-S3A was also used by GlobColour products.

Figure 9. The chlorophyll a concentration (15 December 2017) for (a) the OC-CCI level 3 product and (b) the GlobColour product. provides a homogeneous data set of spectral reflectance that It should be emphasised that this limitation also exists is useful when deriving the chlorophyll a product using a in the sensor chlorophyll a merging approach: in both ap- common algorithm. One could expect better consistency for proaches if a temporal trend is observed it should be carefully the long time series. However, as illustrated above, this as- analysed. sumption is not correct as the homogeneity of the spectral The present findings highlight the advantage of a chloro- reflectance is currently not obtained (spatial and temporal phyll a per sensor merging approach compared with the re- discontinuities exist). flectance merging approach:

www.ocean-sci.net/15/819/2019/ Ocean Sci., 15, 819–830, 2019 828 P. Garnesson et al.: The CMEMS GlobColour chlorophyll a product Based on satellite observation

Figure 10. The chlorophyll a concentration (1 January 2012) for (a) the OC-CCI level 3 product and (b) the GlobColour product. The two initiatives used MODIS Aqua, MERIS and VIIRS-SNPP at this time.

Figure 11. The chlorophyll a concentration (1 January 2012) for (a) the OC-CCI level 3 product and (b) the GlobColour product.

Ocean Sci., 15, 819–830, 2019 www.ocean-sci.net/15/819/2019/ P. Garnesson et al.: The CMEMS GlobColour chlorophyll a product Based on satellite observation 829

– The sensor merging approach facilitates the ingestion two categories: NRT (near-real-time) products covering about of a new sensor or a new reprocessing. Consequently, 1 year of data (extended on a daily basis); and REP (long-time NASA R2018 and OLCI-S3A were successfully intro- series) products starting in 1997. For NRT, the CMEMS catalogue duced in April 2018 for the merged chlorophyll a Glob- references are as follows: Colour chain but are not yet available in the other initia- OCEANCOLOUR_GLO_CHL_L3_NRT_OBSERVATIONS_009 tives (OC-CCI or CMEMS regional product). The ad- _032 (global and daily); OCEANCOLOUR_GLO_CHL_L4_NRT_OBSERVATIONS_009 dition of VIIRS-NOOA20 (JPSS) and OLCI-S3B in the _033 (global, 8 d, monthly and daily-filled); merged GlobColour chlorophyll a product will occur in OCEANCOLOUR_ATL_CHL_L4_NRT_OBSERVATIONS_009 July 2019. _037 (Atlantic and daily-filled). For the REP the CMEMS catalogue references are as follows: – It should be noted that the reflectance merging approach OCEANCOLOUR_GLO_CHL_L3_REP_OBSERVATIONS_009 provides a limited set of six common spectral bands _085 (global and daily); based on the SeaWiFS sensor. For more recent sensors OCEANCOLOUR_GLO_CHL_L4_REP_OBSERVATIONS_009 (i.e. MODIS, VIIRS-SNPP and VIIRS-JPPS1) only five _082 (global, 8 d, monthly and daily-filled); native reflectance bands are available (see Table 1): the OCEANCOLOUR_ATL_CHL_L4_REP_OBSERVATIONS_009 band at 510 nm is obtained by interpolation. Other ex- _098 (Atlantic and daily-filled). Access to products is granted tra bands from MERIS, MODIS or OLCI that are not after free registration at http://marine.copernicus.eu/ (last access: part of the six bands are not usable in the reflectance 15 June 2019). Data can be accessed via different protocols (ftp, WMS and direct access with subsetting capabilities). merging approach (because the spatial complementarity of the sensors cannot be used at a daily level and for the full time series). For the chlorophyll a merging Author contributions. The co-authors are all from the ACRI-ST approach, when extra bands are available (i.e. MERIS, company and are all involved in the in the CMEMS project: PG MODIS and OLCI) they can be used to improve the al- is the technical manager concerning the delivery of the Copernicus- gorithm in the future. This perspective has already been GlobColour products to CMEMS; AM is the product quality leader investigated to retrieve PFT (phytoplankton functional for the CMEMS OCTAC (Ocean Colour Theamtic Assembly Cen- type) from OLCI (e.g. Xi et al., 2018). tre); MB is involved as a scientific expert; JD is the technician re- sponsible of the Copernicus-GlobColour processor; and OFd’A is – The sensor approach provides an improved daily spa- the director of the ACRI-ST company and is the OCTAC consor- tial coverage when OC5 is applied on the sensor re- tium leader for the ACRI-ST PU (production unit). flectance (not on merged reflectance). For the period spanning from 2012 to the present the spatial coverage is improved by an important factor (of about 2.8) Competing interests. The authors declare that they have no conflict when compared with the OC-CCI product. Both open of interest. ocean and coastal areas are improved. This is required for many users involved in the EU Water Framework and Marine Strategy Framework Directive. To satisfy Special issue statement. This article is part of the special is- users interested in coastal data, a better spatial resolu- sue “The Copernicus Marine Environment Monitoring Service (CMEMS): scientific advances”. It is not associated with a confer- tion (300 m) is also required. From this point of view the ence. chlorophyll a merging approach is also more promising (the algorithm can be applied on the level 2 track grid to limit the mixing of the pixels). Acknowledgements. We thank Emmanuel Boss and another anonymous referee for their constructive comments which allowed us to A better spatial coverage is also a key point to help guarantee improve the quality of this paper. the quality of the CMEMS GlobColour chlorophyll a “cloud free” product (referred to as “daily L4” in the CMEMS catalogue) which is based on a spatial and temporal interpolation Review statement. This paper was edited by Pierre-Yves Le Traon of the daily level 3 product. A better daily coverage limits the and reviewed by Emmanuel Boss and one anonymous referee. risk of artefacts due to interpolation.

Data availability. The CMEMS GlobColour chlorophyll products are publicly available on the CMEMS web portal. The GlobColour References chlorophyll products are available as daily, 8 d, monthly and daily ﬁlled (spatial and temporal interpolation) products with a Aiken, J., Moore, G. F., Trees, C. C., Hooker, S. B., and Clark, D. 4 km spatial resolution at the global level and a 1 km resolution K.: The SeaWiFS CZCS-type pigment algorithm, in: SeaWiFS for a European zone called “ATL”. Products are divided into Technical Report Series, edited by: Hooker, S. B. and Firestone, www.ocean-sci.net/15/819/2019/ Ocean Sci., 15, 819–830, 2019 830 P. Garnesson et al.: The CMEMS GlobColour chlorophyll a product Based on satellite observation

E. R., 29, Goddard Space Flight Center, Greenbelt, Maryland, Sathyendranath, S., Grant, M., Brewin, R. J. W., Brockmann, C., USA, 38 pp., 1995. Brotas, V., Chuprin, A., Doerffer, R., Dowell, M., Farman, A., Garnesson, P., Mangin, A., and Bretagnon, M.: Quality User Groom, S., Jackson, T., Krasemann, H., Lavender, S., Martinez Guide, Ocean Colour Production Centre, Satellite Observation Vicente, V., Mazeran, C., Mélin, F., Moore, T. S., Müller, D., GlobColour-Copernicus Products, available at: http://marine. Platt, T., Regner, P., Roy, S., Steinmetz, F., Swinton, J., Valente, copernicus.eu/services-portfolio/access-to-products/?option= A., Zühlke, M., Antoine, D., Arnone, R., Balch, W. M., Barker, com_csw&view=details&product_id=OCEANCOLOUR_ K., Barlow, R., Bélanger, S., Berthon, J.-F., Be¸siktepe, ¸S., GLO_CHL_L3_NRT_OBSERVATIONS_009_032, 15 January Brando, V. E., Canuti, E., Chavez, F., Claustre, H., Crout, R., 2019. Feldman, G., Franz, B., Frouin, R., García-Soto, C., Gibb, S. W., Gohin, F., Druon, J. N., and Lampert, L.: A five channel chlorophyll Gould, R., Hooker, S., Kahru, M., Klein, H., Kratzer, S., Loisel, concentration algorithm applied to SeaWiFS data processed by H., McKee, D., Mitchell, B. G., Moisan, T., Muller-Karger, SeaDAS in coastal waters, Int. J. Remote Sens., 23, 1639–1661, F., O’Dowd, L., Ondrusek, M., Poulton, A. J., Repecaud, 2002. M., Smyth, T., Sosik, H. M., Taberner, M., Twardowski, M., Hu, C., Lee, Z., and Franz, B.: Chlorophyll-a algorithms for olig- Voss, K., Werdell, J., Wernand, M., and Zibordi, G.: ESA otrophic oceans: A novel approach based on three-band re- Ocean Colour Climate Change Initiative (Ocean_Colour_cci): flectance difference, J. Geophys. Res.-Oceans, 117, C01011, Version 3.1 Data, Centre for Environmental Data Analysis, https://doi.org/10.1029/2011JC007395, 2012. https://doi.org/10.5285/9c334fbe6d424a708cf3c4cf0c6a53f5, Le Traon, P. Y., Antoine, D., Bentamy, A., Bonekamp, H., Breivik, 2018. L. A., Chapron, B., Corlett, G., Dibarboure, G., DiGiacomo, Saulquin, B., Gohin, F., and Fanton d’Andon, O.: Inter- P., Donlon, C., and Faugère, Y.: Use of satellite observations polated fields of satellite-derived multi-algorithm chloro- for operational oceanography: recent achievements and future phyll a estimates at global and European scales in the prospects, J. Oper. Oceanogr., 8, s12–s27, 2015. framework of the European Copernicus-Marine Environ- Maritorena, S., d’Andon, O. H. F., Mangin, A., and Siegel, D. A.: ment Monitoring Service, J. Oper. Oceanogr., 12, 47–57, Merged satellite ocean color data products using a bio-optical https://doi.org/10.1080/1755876X.2018.1552358, 2018. model: Characteristics, benefits and issues, Remote Sens. Envi- Steinmetz, F., Deschamps, P.-Y., and Ramon, D.: Atmospheric cor- ron., 114, 1791–1804, 2015. rection in presence of sun glint: application to MERIS, Opt. Morel, A.: Optical properties of oceanic case 1 waters revisited, in: Express, 19, 9783–9800, https://doi.org/10.1364/OE.19.009783, Ocean Optics XIII, International Society for Optics and Photon- 2011. ics, 2963, 108–115 1997. Toole, D. A., Siegel, D. A., Menzies, D. W., Neumann, M. J., and Müller-Karger, F. E., McClain, C. R., Sambrotto, R. N., and Ray, G. Smith, R. C.: Remote-sensing reflectance determinations in the C.: A comparison of ship and coastal zone color scanner mapped coastal ocean environment: impact of instrumental characteris- distribution of phytoplankton in the southeastern Bering Sea, J. tics and environmental variability, Appl. Optics, 39, 456–469, Geophys. Res.-Oceans, 95, 11483–11499, 1990. 2000. O’Reilly, J. E., Maritorena, S., Mitchell, B. G., Siegel, D. A., Vantrepotte, V. and Mélin, F.: Temporal variability of 10-year global Carder, K. L., Garver, S. A., Kahru, M., and McClain, C.: Ocean SeaWiFS time-series of phytoplankton chlorophyll a concentra- color chlorophyll algorithms for SeaWiFS, J. Geophys. Res.- tion, ICES J. Mar. Sci., 66, 1547–1556, 2009. Oceans, 103, 24937–24953, 1998. Volpe, G., Colella, S., Brando, V. E., Forneris, V., La Padula, O’Reilly, J. E., Maritorena, S., Siegel, D. A., O’Brien, M. C., Toole, F., Di Cicco, A., Sammartino, M., Bracaglia, M., Artuso, F., D., Mitchell, B. G., Kahru, M., Chavez, F. P., Strutton, P., Cota, and Santoleri, R.: Mediterranean ocean colour Level 3 op- G. F., Hooker, S. B., McClain, C. R., Carder, K. L., Mueller- erational multi-sensor processing, Ocean Sci., 15, 12700146, Karger, F., Harding, L., Magnusson, A., Phinney, D., Moore, G. https://doi.org/10.5194/os-15-127-2019, 2019. F., Aiken, J., Arrigo, K. R., Letelier, R., Culver, M., Hooker, S. Xi, H., Losa, S. N., Mangin, A., Soppa, M. A., Garnesson, P., B., and Firestone, E. R. (Eds.): Ocean color chlorophyll a algo- Demaria, J., d’Andon, O. H. F., and Bracher, A.: Global Re- rithms for SeawiFS, OC2, and OC, 9–23, Version 4, NASA Tech- trieval Algorithms for Phytoplankton Functional Types (PFTs): nical Memorandum 2000-206892, vol. 20, NASA-GSFC, Green- toward the Applications to OLCI and GlobColour Merged belt, Maryland, USA, 2000. Products, Ocean Optics Conference XXIV, Dubrovnik Croa- Pardo, S., Brando, V., and Taylor, B.: Quality Information Doc- tia, 7 October 2018, hdl:10013/epic.e9a41e28-ef93-43bd-b20a- ument (QUID) of Ocean Monitoring Indicators (OMI) for c613693bd5d8, 2018. OC-TAC Release 1.0, available at: http://marine.copernicus. eu/science-learning/ocean-monitoring-indicators/catalogue (last access: 15 June 2019), 2018.

Ocean Sci., 15, 819–830, 2019 www.ocean-sci.net/15/819/2019/